如何使用NLP和机器学习检测事件？

Question

I have text describing about events such as birth , new job , wedding , death etc .. or no event . How do i detect these events ?

My approach is to form set of words and search them in text corresponding to event . Or use bayesian classifier . But bayesian classifier requires some training on all classes , I require method which can even classify without giving it all types of examples ? Is that possible ?

Answer 1

What are your performance requirements? Is low recall ok? Do you need high precision?

Based on your question, I'm guessing you want something with reasonable recall (reading classified ads to send out spam?) but don't actually have any training data.

You want a method that can classify documents (using events from a Named Entity Recognition algorithm as features) without providing any training data. All supervised methods (including Bayesian ones) require training data, so what you are asking/wishing for is not possible. You need labelled data in any case, otherwise how can you tell how well your detection process is working?

At this stage you should not even be worried about which classifier to use - I suggest writing a handful of regular expressions to see how hard your problem is / what performance you get. It may be a dozen regular expressions may get you 90% of these events and you can avoid over-engineering the problem. Good luck!